Sunday, November 1, 2015

Granularity, predictions, and accuracy - Do you have a scale?


Ask for precision and you get better forecasts - even in an uncertain world. Greater granularity with predictions leads to better thinking. Judgments made on a seven point scale will be better than a five  point scale. Asking for likelihood is better than just asking opinions. Research has found that when forecasters are forced to provide more granularity on their point of view, they will do a better job. (See Superforecasting by Tetlock and Gardner.)

It is always important to get the direction right. One extreme would be trend-followers who focus on direction and not the degree of a price move. However, you should demand more from analysts to better the forecasting process. If given the opportunity, do not let analysts off on ambiguous or imprecise forecasts. If an analyst says bond market yields are going higher, then force a number and level of precision with the forecast. Asking for details will mean more work, but a better forecast in the end. Of course, this does not have to be taken to extremes.

For example, here is a simple scale.

The National Intelligence Council uses either a five or seven point scale for judgements.

The five point scale:
REMOTE - UNLIKELY  - EVEN CHANCE  - PROBABLY, LIKELY  - ALMOST CERTAIN 

The seven point scale:
REMOTE - VERY UNLIKELY  - UNLIKELY  - EVEN CHANCE - LIKELY - VERY LIKELY -ALMOST CERTAIN 


Just by asking for a more refined scale will require more thinking and a better piece of work. So the next time you hear someone say the market is going higher, ask him to give some probability estimate or likelihood to that comment.

No comments: